Autonomous 3D Semantic Mapping of Coral Reefs

EasyChair Preprint 1493

Abstract

This paper presents an approach for autonomous 3D semantic mapping of a coral reef. Current methods on coral reef health assessment are based on human observation. The proposed system organically joins a convolutional neural network (CNN) with a dense visual odometry approach and a correlation filter based tracker, KCF, to identify the different coral species detected during autonomous and manual data collection. In addition to the coral classification, the 3D position of each coral is identified producing a semantic map of the observed reef. Each coral is identified once, even when encountered at different times. The number of different coral species encountered during a single trajectory is reported. Furthermore, the shape and size of each coral are extracted from the dense reconstruction enabling the extraction of volumetric data for subsequent studies. Experimental results from the coral reefs of Barbados, utilizing different vehicles verify the robustness and accuracy of the proposed approach.

Keyphrases: 3D Semantic labeling, Coral habitat mapping, Underwater Robotics, deep learning

@booklet{EasyChair:1493,
  author    = {Md Modasshir and Sharmin Rahman and Ioannis Rekleitis},
  title     = {Autonomous 3D Semantic Mapping of Coral Reefs},
  howpublished = {EasyChair Preprint 1493},
  year      = {EasyChair, 2019}}